As Anthropic debuted on Wall Street Thursday in its largest IPO in history, CEO Dario Amodei reaffirmed the company’s commitment to developing AI responsibly. In a speech before the Nasdaq opening bell, Amodei vowed that AI companies would not bow to either business or political pressure and would “remain under human control.”
Reading a speech from a printed page, the CEO began talking about Anthropic’s origins as a way to build powerful technology to help humanity. He said this was inspired by losing his father to a disease that was now curable just a few years later.
He acknowledged that while human intelligence has brought diseases such as plague, smallpox and polio for most of the world’s history, the number of scientists is still too small and there has been too little time to make advances in other areas of disease. AI could help with this by working with humans to accelerate medical progress, he said.
Amodei has repeatedly warned about the risks of advancing technology too quickly, so it’s telling that he began his talk by focusing on AI’s potential life-saving applications. Amodei made headlines in September when he outlined plans to set the pace for AI development, including coordinating with other major companies on safety standards and incorporating third-party evaluators to verify companies are complying with safety initiatives.
OpenAI CEO Sam Altman, Google DeepMind chief Demis Hassabis and xAI owner Elon Musk said they agreed with Amodei’s position, although few concrete promises were made.
In his speech this morning, Amodei reiterated the need for advances in AI technology and the safety it requires. He cited the “very real risks” of AI in the wrong hands, whether individuals or nation-states, and the potential for “economic disruption” and “loss of control of AI systems.”
He noted that Antropic believes these risks are avoidable as long as safety measures are taken. Amodei said the company will continue to invest in AI tuning and testing work, as well as developing new ways to understand AI behavior so that it can be brought under human control.
“There is still much work to do and the most difficult safety task lies ahead of us,” he said. “The company’s incredible talent, which has enabled us to build some of the most capable AI systems, will now need to be focused more than ever on double-checking that those AI systems are safe and under human control.”
“And let me be clear: We will pace the training and deployment of our AI systems, taking as long as necessary to have confidence that they are safe and under human control,” Amodei concluded.
His sister and Anthropic co-founder Daniela Amodei echoed similar sentiments in an interview with CNBC on Thursday, saying that while the risks are real, the promise of AI is “incredible.”
“We were really grounded in the idea that it’s our job to mitigate these risks, and I think we’ve done a great job of doing that so far. We obviously still have a lot of work to do, but we’re committed to being safe,” she added.
The full text of Dario Amodei’s opening speech is below.
Thank you, Daniela. My co-founders and I founded Anthropic because we believe that the most powerful technologies of our time should be built for the benefit of humanity. Daniela and I lost our father to an illness that he recovered from only a few years later. I started my career in biology with the desire to contribute to the early discovery of future treatments. In the lab, we have seen the enormous complexity that scientists face when trying to understand diseases and develop new treatments. Many of the scientific tools that can cut through that thicket started as pure ideas: tools for gene editing, tools for gene synthesis, and tools for controlling brain cells with light. In other words, they arise from thought and intelligence.
Human intelligence has already written diseases such as plague, smallpox, and polio into history in most parts of the world. But progress in other diseases has come too slowly for too many patients and their families. It can take decades to go from a great idea to a cure. It took many years from the initial discovery of strange repeating patterns in DNA to the approval of the first treatments for disease using CRISPR. Nature is full of these patterns, but our intelligence is too small to decipher most of them. There are too few scientists and too little time. Now we can build that intelligence. Scientists in many areas of biology and medicine, including protein folding, drug discovery, cancer genetics, and clinical trial management, are finding that Claude can help them conduct research in a fraction of the time it took just a few years ago. I believe that with sufficiently powerful AI working in tandem with humans, we can compress a century of scientific and medical progress into a decade.
I don’t want to get ahead of myself. All we have at this point are early results. Although experiments and clinical trials are still time-consuming, time-consuming and labor-intensive processes, we believe that these processes can be significantly accelerated by AI. In the future, we hope that fewer families will miss out on treatment in the coming years. Medicine is just one particularly vivid example, but there are many other applications of Claude. Our customers, the world’s most ambitious businesses, nonprofits, and individuals, are already applying the intelligence we build to stimulate economic growth, lift people out of poverty, protect democracies and free societies, and even improve the functioning of everyday services like the DMV. We are proud to work with them and are very optimistic about what we can do together in the future.
At the same time, we have always felt honestly that there are very real risks with this technology. These range from misuse of technology by individuals or nation states to economic disruption and loss of control of AI systems. We discuss these risks not because we think they are inevitable and are building anyway, but because we think they are avoidable and can be avoided. Avoiding them will depend on the choices of the people building AI. There is a big difference between what AI would do if we paid little attention to safety and what AI would do if we were obsessed with safety. And a focus on safety and the public interest has been in Anthropic’s DNA since the beginning, from safeguards against biological and cyber risks to alignment training and testing, to looking inside models like X-rays, and the company has pioneered ways to understand and manipulate AI behavior and bring it under human control. Much remains to be done and the most difficult safety task lies ahead of us. The company’s incredible talent that has enabled it to build some of the most capable AI systems needs to be focused more than ever on double and triple checking that AI systems are safe and under human control.
And let me be clear: We will pace the training and deployment of our AI systems, taking as long as necessary to have confidence that they are safe and under human control, and we will not bow to either business or political pressure. Our financial position is now stronger, which further strengthens our resolve to do the right thing. If we can show that it is possible to build an incredibly successful company while making the right choices when it comes to safety, we will encourage other companies in our industry to do the same. We call this the race to the top.
When the race to the top is working properly, it doesn’t matter which company has the best model. Everyone wins because everyone can build safely. The race to the top is best achieved by running from the front of the pack, and we really want to do that. But the goal is to make the industry as a whole better, regardless of whether Anthropic is winning or losing at the moment. The emergence of machine intelligence is one of the landmark moments in human history, and we have a unique opportunity to shape it for the better. thank you.
If you make a purchase through links in our articles, we may earn a small commission. This does not affect editorial independence.
